Abnormal User Behavior Detection in Energy Big Data Platform Based on Digital Portrait

With the increasing popularity of energy big data platforms, their security problems of insider threats have attracted a lot of attentions. Abnormal behavior detection based on user digital profiles is an effective way to prevent such threats. Existing approaches underutilize priori knowledge about these platforms and mainly focus on improving the accuracy of their algorithms. In this paper, we propose a digital portrait construction method based on multi-model fusion for abnormal user behavior detection. The method is based on the behavioral characteristics of users who accesses energy big data platforms. Specially, LSPRFIA is proposed to model the discrete features of user behaviors, and MTWLSTMIA is proposed to model the continuous features. Finally, LSMT model is formed by combing LSPRFIA with MTWLSTMIA, which improves the accuracy rate for detecting insider attacks. Moreover, an abnormal behavior detection framework is constructed to further boost the detection accuracy for different types of platform users. Experimental results show that the proposed method can be effectively applied to energy big data platforms with high accuracy of abnormal behavior detection.